This repository is considered EXPERIMENTAL and under active development until further notice. It is subject to non-backward compatible changes or removal in any future version.

• Overview
• Key Advantages
  • Performance
  • Security
  • Isolation & Reliability
• Usage
  • Build reusable workflow
  • Bake reusable workflow

This repository provides official Docker-maintained reusable GitHub Actions workflows to securely build container images and artifacts using Docker best practices. The reusable workflows incorporate functionality from our GitHub Actions like build-push-action, login-action, metadata-action, etc., into a single workflow:

name: ci

permissions:
  contents: read

on:
  push:
    branches:
      - 'main'
    tags:
      - 'v*'
  pull_request:

  build:
    uses: docker/github-builder-experimental/.github/workflows/build.yml@main
    permissions:
      contents: read # to fetch the repository content
      id-token: write # for signing attestation(s) with GitHub OIDC Token
    with:
      output: image
      push: ${{ github.event_name != 'pull_request' }}
      meta-images: name/app
    secrets:
      registry-auths: |
        - registry: docker.io
          username: ${{ vars.DOCKERHUB_USERNAME }}
          password: ${{ secrets.DOCKERHUB_TOKEN }}

This workflow provides a trusted BuildKit instance and generates signed SLSA-compliant provenance attestations, guaranteeing the build happened from the source commit and all build steps ran in isolated sandboxed environments from immutable sources. This enables GitHub projects to follow a seamless path toward higher levels of security and trust.

• Native parallelization for multi-platform builds.
  Workflows can automatically distribute builds across runners based on target platform to be built, improving throughput for other architectures without requiring emulation or custom CI logic or self-managed runners.

• Optimized cache warming & reuse.
  The builder can use the GitHub Actions cache backend to persist layers across branches, PRs, and rebuilds. This significantly reduces cold-start times and avoids repeating expensive dependency installations, even for external contributors' pull requests.

• Centralized build configuration.
  Repositories no longer need to configure buildx drivers, tune storage, or adjust resource limits. The reusable workflows encapsulate the recommended configuration, providing fast, consistent builds across any project that opts in.

• Trusted workflows in the Docker organization.
  Builds are executed by reusable workflows defined in the @docker organization, not by arbitrary user-defined workflow steps. Consumers can rely on GitHub's trust model and repository protections on the Docker side (branch protection, code review, signing, etc.) to reason about who controls the build logic.

• Verifiable, immutable sources.
  The workflows use the GitHub OIDC token and the exact commit SHA to obtain source and to bind it into SLSA provenance. This ensures that the build is tied to the repository contents as checked in—no additional CI step can silently swap out what is being built.

• Signed SLSA provenance for every build.
  BuildKit generates SLSA-compliant provenance attestation artifacts that are signed with an identity bound to the GitHub workflow. Downstream consumers can verify:

  • which builder commit produced the image
  • which source code commit produced the image
  • which workflow and job executed the build
  • what inputs and build parameters were used

• Protection from user workflow tampering.
  The build steps are pre-defined and optimized in the reusable workflow, and cannot be altered by user configuration. This protects against tampering: preventing untrusted workflow steps from modifying build logic, injecting unexpected flags, or producing misleading provenance.

• Separation between user CI logic and build logic.
  The user's workflow orchestrates when to build but not how to build. The actual build steps live in the Docker-maintained reusable workflows, which cannot be modified from the consuming repository.

• Immutable, reproducible build pipeline.
  Builds are driven by declarative inputs (repository commit, build configuration, workflow version). This leads to:

  • reproducibility (same workflow + same inputs → same outputs)
  • auditability (inputs and workflow identity recorded in provenance)
  • reliability (less dependence on ad-hoc per-repo CI scripting)

• Reduced CI variability and config drift.
  By reusing the same workflows, projects avoid maintaining custom build logic per repository. Caching, provenance, SBOM generation, and build settings behave uniformly across all adopters.

• Higher assurance for downstream consumers.
  Because artifacts are produced by a workflow in the @docker organization, with SLSA provenance attached, consumers can verify both the source commit and the builder identity before trusting or promoting an image, an essential part of supply-chain hardening.

name: ci

permissions:
  contents: read

on:
  push:
    branches:
      - 'main'
    tags:
      - 'v*'
  pull_request:

  build:
    uses: docker/github-builder-experimental/.github/workflows/build.yml@main
    permissions:
      contents: read # to fetch the repository content
      id-token: write # for signing attestation(s) with GitHub OIDC Token
    with:
      output: image
      push: ${{ github.event_name != 'pull_request' }}
      meta-images: name/app
      meta-tags: |
        type=ref,event=branch
        type=ref,event=pr
        type=semver,pattern={{version}}
      build-platforms: linux/amd64,linux/arm64
    secrets:
      registry-auths: |
        - registry: docker.io
          username: ${{ vars.DOCKERHUB_USERNAME }}
          password: ${{ secrets.DOCKERHUB_TOKEN }}

  # Optional job to verify the pushed images' signatures. This is already done
  # in the `build` job and can be omitted. It's provided here as an example of
  # how to use the `verify.yml` reusable workflow.
  build-verify:
    uses: docker/github-builder-experimental/.github/workflows/verify.yml@main
    if: ${{ github.event_name != 'pull_request' }}
    needs:
      - build
    with:
      builder-outputs: ${{ toJSON(needs.build.outputs) }}
    secrets:
      registry-auths: |
        - registry: docker.io
          username: ${{ vars.DOCKERHUB_USERNAME }}
          password: ${{ secrets.DOCKERHUB_TOKEN }}

You can find the list of available inputs in .github/workflows/build.yml.

name: ci

permissions:
  contents: read

on:
  push:
    branches:
      - 'main'
    tags:
      - 'v*'
  pull_request:

  bake:
    uses: docker/github-builder-experimental/.github/workflows/bake.yml@main
    permissions:
      contents: read # to fetch the repository content
      id-token: write # for signing attestation(s) with GitHub OIDC Token
    with:
      output: image
      push: ${{ github.event_name != 'pull_request' }}
      meta-images: name/app
      meta-tags: |
        type=ref,event=branch
        type=ref,event=pr
        type=semver,pattern={{version}}
    secrets:
      registry-auths: |
        - registry: docker.io
          username: ${{ vars.DOCKERHUB_USERNAME }}
          password: ${{ secrets.DOCKERHUB_TOKEN }}

  # Optional job to verify the pushed images' signatures. This is already done
  # in the `bake` job and can be omitted. It's provided here as an example of
  # how to use the `verify.yml` reusable workflow.
  bake-verify:
    uses: docker/github-builder-experimental/.github/workflows/verify.yml@main
    if: ${{ github.event_name != 'pull_request' }}
    needs:
      - bake
    with:
      builder-outputs: ${{ toJSON(needs.bake.outputs) }}
    secrets:
      registry-auths: |
        - registry: docker.io
          username: ${{ vars.DOCKERHUB_USERNAME }}
          password: ${{ secrets.DOCKERHUB_TOKEN }}

You can find the list of available inputs in .github/workflows/bake.yml.